Re: [Pytables-users] numpy datatypes

[Jan S. <cur...@gm...>]

Hi Francesc,
I'm sorry, but this fix still doesn't do it for me.
It now runs without messages, but the result is different from casting to
long.
In other words running table.where(col1 == rumber) runs, but the result is
different from table.where(col1 == long(number))

I don't have a good testcase handy (my case is 17M), but have you considered
unsigned types as well ?
When I have some time I try and reduce the test case to something more
manageable.
Thanks,
    Jan

On 1/11/07, Jan Strube <cur...@gm...> wrote:
>
> Hi !
> I have a problem with the numpy datatypes and the Table.where method.
> something like
> for item in numpy_list:
>    print [row for row in table.where(table.cols.item==item)]
> fails, if numpy_listis a list of numpy integers.
> Traceback (most recent call last):
>   File "createDisjointSets.py", line 101, in ?
>     createDisjointSets(sys.argv[1])
>   File "createDisjointSets.py", line 90, in createDisjointSets
>     copyTable( fLambda.root.events.eventHeader, events, uniqueLambdas)
>   File "createDisjointSets.py", line 62, in copyTable
>     rows = [row.nrow for line in selector for row in source.where(
> source.cols.event_id == line['event_id'])]
>   File
> "/nfs/slac/g/lcd/mc/prj/sw/extern/python2.4/lib/python2.4/site-packages/tables/Table.py",
> line 763, in where
>     return self._whereIndexed(condition, start, stop, step)
>   File
> "/nfs/slac/g/lcd/mc/prj/sw/extern/python2.4/lib/python2.4/site-packages/tables/Table.py",
> line 823, in _whereIndexed
>     ncoords = condition.index.getLookupRange(condition)
>   File
> "/nfs/slac/g/lcd/mc/prj/sw/extern/python2.4/lib/python2.4/site-packages/tables/Index.py",
> line 607, in getLookupRange
>     raise TypeError("""\
> TypeError: Bounds (or range limits) for integer columns can only be ints
> or floats.
>
> I can make it work by casting item to an int, but I really don't want to.
> (In some cases I can't)
> The flavor of the table is numpy and the size (in bytes) of the datatypesis the same
> Is this a bug, or am I just doing something silly and there is a better
> way to do things?
>
> Thanks,
>     Jan
>